The increased sensitivity of the updated circuit means you can use a ready-packaged power-darlington as the 'antitriode' with almost no changes.
The old BU808DFI (TV sweep darlington) will run less than 90uA into the base, and will yield such high gm that the input impedance is effectively zero. SOA is no problem at 250V, even in an isolated package.
Some of the ignition-coil darlingtons might also be worth a look [ST microel, etc]
The old BU808DFI (TV sweep darlington) will run less than 90uA into the base, and will yield such high gm that the input impedance is effectively zero. SOA is no problem at 250V, even in an isolated package.
Some of the ignition-coil darlingtons might also be worth a look [ST microel, etc]
I already mentioned that Don offered a solution for this anomaly for pentode-tetrode connected finals.
I have taken little interest in this aspect because my own version of this design (which I am now beginning to like a good deal) will be a larger amp - 6L6Gx triode connected AB2 with toroidal OTX. The cost of toroids doesn't increase much between 50VA and 300VA, for my preferred vendor at least. And the circuit is already showing a Power-FET grid driver; we may as well make it work for its living.
I do hope you like the vanished electrolytic version.
I have taken little interest in this aspect because my own version of this design (which I am now beginning to like a good deal) will be a larger amp - 6L6Gx triode connected AB2 with toroidal OTX. The cost of toroids doesn't increase much between 50VA and 300VA, for my preferred vendor at least. And the circuit is already showing a Power-FET grid driver; we may as well make it work for its living.
I do hope you like the vanished electrolytic version.
Rod,
Minor detail fix on schematic of post #58. Without Q4 in the circuit, R20 needs to go over to the R14 side along with R16 too.
"The cost of toroids doesn't increase much between 50VA and 300VA"
I assume this is something oriented toward audio with enough primary inductance. The ordinary power toroidals drop in inductance with bigger size due to fewer turns.
"6L6Gx triode connected AB2 with toroidal OTX"
With AB mode, the cross connected cathode to slave/anti-triode scheme won't work. Assuming this would mean conventional P-P AB2 design with toroid OT. Although there is the untried, so far, idea of using the anti-triode scheme for a driver stage, which could then drive a conventional P-P stage in ABx mode.
Ken,
Your Blumlein derived garter bias circuit probably can be made to track quite well versus temp. if the two bipolar transistors are a matched pair. Analog Devices makes some dual matched parts. MAT01, MAT02 ...MAT04, and SSM-1 .... as I recall.
Don
Minor detail fix on schematic of post #58. Without Q4 in the circuit, R20 needs to go over to the R14 side along with R16 too.
"The cost of toroids doesn't increase much between 50VA and 300VA"
I assume this is something oriented toward audio with enough primary inductance. The ordinary power toroidals drop in inductance with bigger size due to fewer turns.
"6L6Gx triode connected AB2 with toroidal OTX"
With AB mode, the cross connected cathode to slave/anti-triode scheme won't work. Assuming this would mean conventional P-P AB2 design with toroid OT. Although there is the untried, so far, idea of using the anti-triode scheme for a driver stage, which could then drive a conventional P-P stage in ABx mode.
Ken,
Your Blumlein derived garter bias circuit probably can be made to track quite well versus temp. if the two bipolar transistors are a matched pair. Analog Devices makes some dual matched parts. MAT01, MAT02 ...MAT04, and SSM-1 .... as I recall.
Don
Last edited:
With toroidals is a toss up between to little core to handle the low frequencies, and not enough primary inductance for the same frequencies.
My personal experience with the Telama's brand is that response is good down to 20hz with 150VA transformers. I have to say though that all my designs have been Shade or 6080 based so they all have very low output impedance which makes them well able to bite on the available inductance.
Shoog
My personal experience with the Telama's brand is that response is good down to 20hz with 150VA transformers. I have to say though that all my designs have been Shade or 6080 based so they all have very low output impedance which makes them well able to bite on the available inductance.
Shoog
Hey Rod,
Couldn´t keep away from the FET s
. Also removed the "anode-caps" and returned to the 33 ohms. Seems to work quite well in theory(LTSpice) when pushing the ECL86 both up and down quite a bit. Forgot to add gatestoppers for the PFETs:
About the bigger toroids you are looking at, they seem to have low inductance over 80VA. Have you measured?
Couldn´t keep away from the FET s
. Also removed the "anode-caps" and returned to the 33 ohms. Seems to work quite well in theory(LTSpice) when pushing the ECL86 both up and down quite a bit. Forgot to add gatestoppers for the PFETs:An externally hosted image should be here but it was not working when we last tested it.
About the bigger toroids you are looking at, they seem to have low inductance over 80VA. Have you measured?
Lars, you certainly like your FETs.
Only thing to watch is that the sample-to-sample variation of FETs is much worse than BJTs, so the balance will be worse - you may need a trimmer to centre the diffl amp. you can build the rest of the circuit the same if you substitute the MPSA92s again - with a current source tail, the input impedance problem vanishes.
As for toroids, I would be interested in recommendations for mains parts - thanks for yours, Shoog.
Only thing to watch is that the sample-to-sample variation of FETs is much worse than BJTs, so the balance will be worse - you may need a trimmer to centre the diffl amp. you can build the rest of the circuit the same if you substitute the MPSA92s again - with a current source tail, the input impedance problem vanishes.
As for toroids, I would be interested in recommendations for mains parts - thanks for yours, Shoog.
Selecting a power toroid for OPT use seems to be a matter of some trial and error. I find the balance is between DCR and inductance, as Don says. The smaller ones have more of both. Also, one needs to measure with various connections and polarities to minimize the impact of capacitance. Some models will just not work as well as others and it's not easy to predict.
One consideration is that the AC saturation voltage hence maximum output power will be pretty much the same for a given primary AC voltage rating regardless of VA rating. Going to a bigger toroid mainly reduces DCR at the expense of less Lpri...
Screen current may not be a huge issue for many tubes at modest plate voltage swing. It would add some distortion but maybe it wouldn't sound bad.
If you go A2, you will definitely want the G1 current out of the plate feedback current loop, and A2 with beam tubes can result in high screen current.
I would probably shoot for <1mA DC offset using a toroid but I haven't actually tried to measure LF distortion at various DC currents on any toroids.
I don't see what is wrong with an adjustment.
One consideration is that the AC saturation voltage hence maximum output power will be pretty much the same for a given primary AC voltage rating regardless of VA rating. Going to a bigger toroid mainly reduces DCR at the expense of less Lpri...
Screen current may not be a huge issue for many tubes at modest plate voltage swing. It would add some distortion but maybe it wouldn't sound bad.
If you go A2, you will definitely want the G1 current out of the plate feedback current loop, and A2 with beam tubes can result in high screen current.
I would probably shoot for <1mA DC offset using a toroid but I haven't actually tried to measure LF distortion at various DC currents on any toroids.
I don't see what is wrong with an adjustment.
Last edited:
"Also, one needs to measure with various connections and polarities to minimize the impact of capacitance. Some models will just not work as well as others and it's not easy to predict."
I've found that some of these toroids have the two 120V windings wound bifilar, which makes for huge capacitance problems when connected in series. Check common mode C with a C meter between the two windings with each winding shorted. If it needs the nF scale, like 20 nF, it probably is bifilar wound. 600 pF is more satisfactory. Has anyone checked those Antek audio OT toroids?
Don
I've found that some of these toroids have the two 120V windings wound bifilar, which makes for huge capacitance problems when connected in series. Check common mode C with a C meter between the two windings with each winding shorted. If it needs the nF scale, like 20 nF, it probably is bifilar wound. 600 pF is more satisfactory. Has anyone checked those Antek audio OT toroids?
Don
Last edited:
Member
Joined 2009
Paid Member
This is great. I am very interested in SE Triode without an air-gapped transformer. I will study this further....
"This is great. I am very interested in SE Triode without an air-gapped transformer."
There are some other ways too:
inverted triode feedback:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=20686&highlight=
(Hmm, looks like the diagram has gotten inverted B/W-wise with the new forum software. The triode plate connects to the amplifier output thru a cap, the grid gets grounded, and the cathode is CCS loaded and connected to the inverting input of the amplifier, the inverted triode acting as the amplifier feedback attenuator network)
(some related schemes on the Web, check here:
http://www.ne.jp/asahi/evo/amp/el86/report.htm)
V and I composite amplifiers:
http://www.diyaudio.com/forums/showpost.php?p=595103&postcount=44
(Probably not the scheme used by the commercial design mentioned in that post, but the diagram at least illustrates the composite idea. The tube/SS Darlington/ CFP approaches are also of this type. Main idea is for the current amplifier to unload the voltage amplifier, but the voltage amplifier controls the voltage)
mixed class amplifiers:
http://www.diyaudio.com/forums/showpost.php?p=1409048&postcount=3
(this is really just another form of composite amplifier, but V amp + V amp summing mode here. The tube amp puts the final touch on the sound by means of feedback control)
probably some others? (well, of course, there is also the venerable para-feed approach)
But the anti-triode scheme has the advantage of being able to closely mimick the full internals of a SE amp design (input, driver stages), and a hopefully accurate triode emulating output stage. (the present all pentode/Schading/ Mosfet scheme is stretching this to the limit, but looks hopeful nevertheless)
Don
There are some other ways too:
inverted triode feedback:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=20686&highlight=
(Hmm, looks like the diagram has gotten inverted B/W-wise with the new forum software. The triode plate connects to the amplifier output thru a cap, the grid gets grounded, and the cathode is CCS loaded and connected to the inverting input of the amplifier, the inverted triode acting as the amplifier feedback attenuator network)
(some related schemes on the Web, check here:
http://www.ne.jp/asahi/evo/amp/el86/report.htm)
V and I composite amplifiers:
http://www.diyaudio.com/forums/showpost.php?p=595103&postcount=44
(Probably not the scheme used by the commercial design mentioned in that post, but the diagram at least illustrates the composite idea. The tube/SS Darlington/ CFP approaches are also of this type. Main idea is for the current amplifier to unload the voltage amplifier, but the voltage amplifier controls the voltage)
mixed class amplifiers:
http://www.diyaudio.com/forums/showpost.php?p=1409048&postcount=3
(this is really just another form of composite amplifier, but V amp + V amp summing mode here. The tube amp puts the final touch on the sound by means of feedback control)
probably some others? (well, of course, there is also the venerable para-feed approach)
But the anti-triode scheme has the advantage of being able to closely mimick the full internals of a SE amp design (input, driver stages), and a hopefully accurate triode emulating output stage. (the present all pentode/Schading/ Mosfet scheme is stretching this to the limit, but looks hopeful nevertheless)
Don
Last edited:
Member
Joined 2009
Paid Member
My 'borg' brain has assimilated a dozen threads on this - not sure how effectively though. Wow, this is great, something just facinates me about SE triodes and their hybrid Triode-SS cousins. And I've never heard one.
I'm not sure yet which ideas are the most compelling. As I see it, we'd like to have
i) SE output sound
ii) greater efficiency (choke loaded if SE, otherwise PP)
iii) good power output (not limited by need of an air-gapped transformer)
But so far the ideas I've seen offer only 2 out of 3 of the above desires ?
I'm not sure yet which ideas are the most compelling. As I see it, we'd like to have
i) SE output sound
ii) greater efficiency (choke loaded if SE, otherwise PP)
iii) good power output (not limited by need of an air-gapped transformer)
But so far the ideas I've seen offer only 2 out of 3 of the above desires ?
Well, this works out pretty good, till I try CCS shunt reg screen.
Then LTSpice can't seem to find stable operating point to start.
Just want to throw about 8mA at each leg, what don't care
how much goes thru a pentode screen or shunt reg past it...
Adjust Garters back up to 40mA per leg (counting the extra 8).
I think some days its just this sim too stOOpid to let me do stuff
that I'm sure works fine in real life..
Then LTSpice can't seem to find stable operating point to start.
Just want to throw about 8mA at each leg, what don't care
how much goes thru a pentode screen or shunt reg past it...
Adjust Garters back up to 40mA per leg (counting the extra 8).
I think some days its just this sim too stOOpid to let me do stuff
that I'm sure works fine in real life..
Attachments
Last edited:
What's missing?
i) SE output sound. Anti-triode P-P has this... plus
ii) greater efficiency. To the limit of class A operation... plus
iii) good power output. as good as any class A P-P and not needing gapped transformers, although I think a small gap as many P-P OPTs have is helpful for the slight DC shift caused by the SE signature itself (f2 added at high signal level). Even if you use a servo balancer it has a time constant.
The limit I hit has more to do with class A operation and how much heat you want a semiconductor part to dissipate. The FET-tube cascode helps a little.
I'm working on getting (i) with even more of (ii) and (iii) using class AB, but that's not even experimental yet.
Ken,
Very interesting design. Looks like V2 cathode is slaved to U1,U2 plate via M1 feedback, then M4 anti-trioded to V2. And M1 provides the hi Zo to drive the Schade feedback point for V2.
Could get rid of the last triode (U1,U2) and put a pentode there. Plate as output to the Schade point in place of M1, and screen grid as input from V2 cathode (buffered). Just an idea. No obvious advantage.
Michael,
Class AB output certainly raises the bar for difficulty. Only obvious way I can see would be using the anti-triode scheme for a driver/splitter, followed by a very transparent class AB stage. Maybe error correction or some kind of local feedback for the AB stage. Likely also need some feedback to the driver stage "triode" like Ken has in his recent schematic. Making the whole output stage a slave to the triode.
Or, an anti-triode scheme for a driver/splitter but using pentodes (or modulated cascodes as Ken used), then the class AB P-P outputs would individually feed back to the driver tube screen grids. CCS pullups on the driver plates to get high loop gain between each driver and output tube. Effectively slaving each P-P side to each driver tube individually.
Bigun,
Yep, class A is just not efficient. Notice how many of the posters in this thread hail from cool climates?
Don
Very interesting design. Looks like V2 cathode is slaved to U1,U2 plate via M1 feedback, then M4 anti-trioded to V2. And M1 provides the hi Zo to drive the Schade feedback point for V2.
Could get rid of the last triode (U1,U2) and put a pentode there. Plate as output to the Schade point in place of M1, and screen grid as input from V2 cathode (buffered). Just an idea. No obvious advantage.
Michael,
Class AB output certainly raises the bar for difficulty. Only obvious way I can see would be using the anti-triode scheme for a driver/splitter, followed by a very transparent class AB stage. Maybe error correction or some kind of local feedback for the AB stage. Likely also need some feedback to the driver stage "triode" like Ken has in his recent schematic. Making the whole output stage a slave to the triode.
Or, an anti-triode scheme for a driver/splitter but using pentodes (or modulated cascodes as Ken used), then the class AB P-P outputs would individually feed back to the driver tube screen grids. CCS pullups on the driver plates to get high loop gain between each driver and output tube. Effectively slaving each P-P side to each driver tube individually.
Bigun,
Yep, class A is just not efficient. Notice how many of the posters in this thread hail from cool climates?
Don
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.